1380P - The quality of sample size calculation (SSC) reporting in cancer clinical trials

Date 30 September 2012
Event ESMO Congress 2012
Session Poster presentation II
Topics Pathology/Molecular Biology
Presenter Giovanni Bariani
Authors G.M. Bariani1, A. Ferrari2, P.M. Hoff3, R. Arai3, M. Precivale4, R.P. Riechelmann3
  • 1Clinical Oncology, Instituto do Cancer do Estado de Sao Paulo, 01246-000 - Sao Paulo/BR
  • 2Instituto do Cancer do Estado de Sao Paulo, 01246-000 - Sao Paulo/BR
  • 3Medical Oncology, Instituto do Cancer do Estado de Sao Paulo, 01246-000 - Sao Paulo/BR
  • 4Statistics, PGS Statistics, 05418 000 - Sao Paulo/BR

Abstract

Background

SSC is a pivotal step in clinical trial concept and design. It determines the chance of detecting a significant result, ensures appropriate power, and helps sponsors to allocate adequate resources into trials. Here we describe the frequency with which randomized cancer clinical trials (RCT) report the data required for SSC.

Methods

We systematically searched for phase III RCT published in top clinical oncology journals which were accompanied by editorials from Jan 2008 to Oct 2011. We assumed that RCT discussed by editorialists were clinically important. Two blinded investigators extracted data on SSC. Table 1 describes the information required for SSC according to variable type used for primary endpoint.

Results

140 out of 150 RCT were eligible. Median sample size was 596 subjects (50-40,000) per RCT. In 65.7%; of RCT, the number of enrolled subjects was at least 90%; of the planned sample size. The primary endpoint was a categorical variable in 10.0%;, continuous in 27.9%;, and time-to-event in 62.1%;. In general, 80.7%; reported a planned sample size, 57.9%; described their null hypothesis (H0), with 20.7% giving a scientific rationale for H0. 57.9%; informed their alternative hypothesis (H1). Alpha (α) and beta (β ) errors were explicit in 92.9%; and 90.7%;, respectively. Expected difference between arms was reported in 88.6%;. Only 2.9%; of RCT provided all information for proper SSC (required and optional, Table). Excluding “optional information”, SSC could be reproducible in 18.6%; of RCT.

Conclusion

Regardless of the CONSORT 2010 statement, the quality of SSC reporting in phase III cancer RCT seems inadequate. This may compromise future study designs, pooling of data and interpretation of results. Lack of transparency in SSC reporting may also have ethical implications.

Information necessary for SSC

Variable Type Required information Optional information
Categorical H1H0Expected difference between arms (delta) α and β errors Dropout rateStatistical test used for SSCScientific rationale for H0 and H1
Continuous H1H0DeltaStandard deviation for both arms α and β errors Same as above
Time-to-event H1H0Delta (hazard ratio)Length of recruitmentLength of follow up for each patient α and β errors Same as above
Disclosure

G.M. Bariani: Travel expenses covered: Novartis, Roche.

A.C.R.C. Ferrari: Travel expenses covered: Roche.

R.P. Riechelmann: Honoraria (Roche, Merk Serono, Novartis and Bayer). Consultancy (Novartis and Merck Serono). Travel to scientific meetings (Roche, Merk Serono, Novartis e Bayer).

All other authors have declared no conflicts of interest.